
you actually just have to solve real-world AI
The word "just" is working very hard in that sentence
The UK government has confirmed planned revisions to the Highway Code to accommodate self-driving vehicles, including allowing drivers to watch TV while an AI takes the wheel. In a moment history may judge as legislative hubris, the Department for Transport (DfT) said the modifications would include "allowing drivers to view …
Seems to me having the intelligence in the car is a waste of time. We need to have intelligence in the roads. If instead of trying to get AI to read road markings and traffic lights we had a system that unequivocally communicated to the cars if they could stop or go or what speed they could travel at and in which lane and, importantly, on which roads assisted driving was safe, we would have partial self driving a lot quicker. Its the boring bit of the journey that is easiest to automate.
Re "illegal to use mobile phones in self-driving mode, given the greater risk they pose in distracting drivers as shown in research."
I think we will find in research in the near future that shows watching TV while driving is similarly distracting.
Probably from a misinformed Tesla driver within days of this legislation being past.
you actually just have to solve real-world AI
That's assuming that:
1. Real world AI is a solvable problem.
2. An autonomous vehicle with judgement as bad as that of many humans -- a very likely result of AI that "works" -- is acceptable.
3. Manufacturers will not be sued for a zillion dollars every time their product kills or maims a human being.
Personally, I don't think Clippy really ought to be given a driving license any time soon. Maybe not any time ever.
AI will never be as good as a human, while it relies on sensors. A while ago, I was driving to work, and a van ahead of me was pulling off some dumbass 3 point turn at a junction. I saw something move through the van's wheels (road was on an incline) so I slowed down, and then the vehicle behind the van pulled out of the junction, despite not being able to see it was clear. AI will not make use of clues such as partial views of vehicles, it will not see reflections and realise a vehicle is present. it will not see reflections of headlights and work out vehicles are present. Even if other FSD cars had transponders to show where they were, pedestrians, cyclist and those darned e-scooters won't.
You make a good point about sensors, but a wider question is about just what sensors should we be using.
Visual (cameras); lots of them but have numerous drawbacks. Degraded performance in poor visibility being the obvious one. In the USA (or any other place that can get quite hot) the rising air from the tarmac distorts the image significantly. Whether that is a problem or not is not something I know.
Lidar. Performance degrades in wet weather (rain circularly polarises electromagnetic radiation). At short ranges probably not an issue but thick fog might be a different kettle of fish.
mm Wave RF. Suffers significant attenuation depending on weather and is susceptible to multipath fading. Even using beamforming techniques, there is still quite a wide -3dB angle and sidelobes need to be taken into account. Anything beyond a few metres might be problematic.
A mixture of all of them might be better but then there would also need to be redundancy (how many people get a duff brake light fixed? I expect broken sensors would get the same level of attention).
Then there is the matter of stitching all that data together to get a coherent 'view' of the surroundings.
My take is there is a long way to go (and good luck on the country lanes in Cornwall and many other places in the UK).
I have to drive a minimum of 1.5 miles (depending on which direction I leave the village) to get to a 'proper' road and even the road to the village is a single track lane. I am not sure the existing level of processing could figure out what to do if another vehicle suddenly pulls into view (sometimes it is necessary to reverse a bit for a suitable 'passing' area).
but that probably puts you in the top 1% of drivers who even noticed that and could react to it, after all it was a "human" who was making the dumbass 3 point turn, and a "human" who pulled out of the junction despite not being able to see it was clear.
and what you are demanding from AI is perfection, whilst actually accepting far from perfection from your superior to AI humans you share the roads with currently
there were 1460 people killed in road collisions during 2020, during a Covid imposed national lockdown. thats 4 people per day killed on the UKs roads, whilst most people actually stayed at home. 23,529 killed/seriously injured, 115,584 of all injury severities, all involving superior human driving skills.
if AI reduces those numbers even by a fraction, it will be far superior than humans at driving, even if it never achieves perfection.
if AI reduces those numbers even by a fraction
"If". That's a heavy load for a small word to carry. Consider the number of vehicle miles travelled during that time. When you work miles travelled per accident you're setting the AI a tough challenge.
"If". Aye, there's the rub.
Self driving cars need to be safer than human drivers. That is not a high bar and will be easily achievable. Humans are appalling drivers.
However, the existing roads and Highway Code are set up for humans and they will need to be adapted to accommodate the self driving cars, rather than trying to make the self driving cars adapt to human environments.
You say that driving better than a human will be easily achievable. They were saying that ten years ago and haven't managed it yet.
Adapting the roads and highway code to accomodate self driving cars is just admitting that self driving cars can't do the same job as a human.
I think a decent step in the mean time is for assisted driving vehicles to be able to communicate to each other. Ships do this. On long straight roads the vehicle in front can tag along, with the driver (or autonomous vehicle) in the front handling obstructions up ahead. Between long stretches of motorway with limited exits this would almost certainly be safer than each driver having to stay alert. The vehicle in the front is closer to the obstruction. You could achieve near instantaneous simultaneous braking and avoid the concertina effect. Without the need for all cars to be autonomous.
Re-read your last sentence - what a dystopian view! Roads are for humans. How are you going to adapt them to be 'safe' for dumb robots? The only practical way would be to put fences on each side and limit human access to designated zones where the robot cars were stationary. You could call them something like 'stations'. Then you need to be 100% sure the robots stay on path - better put a mechanical fail safe to back up the electronics - you could probably just get the cars to follow a couple of metal strips in the road.
Meanwhile in the real world I quite like walking and cycling on roads, and being a selfish git I value my freedom to do that more than I value some other selfish git's precious need to catch up on Love Island on his way to a pointless meeting about telephone sanitizing
Motorways are already dedicated to motor vehicles capable of some minimum performance.
Which is why motorways are very likely going to be the only places full-self-driving might happen within the next 50 years.
Anyone is allowed on dual carriageways, despite the speed limit often being the same.
"Self driving cars need to be safer than human drivers. That is not a high bar and will be easily achievable. Humans are appalling drivers."
You'd be surprised how high that bar actually is. Even though it doesn't seem like it, the AVERAGE human driver is pretty good and pretty safe. The assholes stick out like a sore thumb and you remember those. Not the millions of other vehicles you encounter a year that just pass by without notice. Humans are relatively good at dealing with edge cases and "new" scenarios. We can infer lots of possible scenarios and anticipate for the majority of them in near real-time. AI is currently failing at exactly those sorts of things. It doesn't anticipate, it doesn't imagine or think ahead.
You'd be surprised how high that bar actually is. Even though it doesn't seem like it, the AVERAGE human driver is pretty good and pretty safe. The assholes stick out like a sore thumb and you remember those.
But it only takes one asshole and there are hundreds of thousands of them. Just look at how many Audis there are on the road.
I'd like to up vote you and down vote you at the same time.
Self-driven cars is hard, especially in the last mile. Factors like, it is pub kicking out time the pedestrians are behaving erratically, are probably never going to better handled by computers. Factors like the speed limit is 50 are probably better handled by computers.
Decisions like it is safer to go faster than the speed limit right now are potentially impossible for computers.
"Decisions like it is safer to go faster than the speed limit right now are potentially impossible for computers."
But thats a decision where there is always one answer - NO. Thats not even a binary question.
A "self driving" car that accelerated and braked smoothly and always obeyed the speed limits would be a quantum leap forward both for Safety and Climate change.
OK, try the opposite. When is it not safe to drive at the speed limit? That's one that calls the whole issue of speed limits into question. "We" set speed limits on the assumption that we can do better than the driver on the spot and yet in adverse conditions "we" expect the driver on the spot to set a safe speed that's lower than the speed limit. So what's the basis for that assumption?
In any difficult situation relating to driving it seems that it's the driver, on the spot, seeing the situation, feeling how the car responds to steering and breaking input, that is to be relied on to make the best decision.
"a quantum leap forward both for Safety and Climate change."
If by "quantum leap" you mean a step change and not a small one, then the best option is to rejig the way we work to reduce communing, We've just demonstrated that that is possible so why not make it permanent?
> However, the existing roads and Highway Code are set up for humans and they will need to be adapted to accommodate the self driving cars, rather than trying to make the self driving cars adapt to human environments.
I'd be up for this. England has some truly appalling road designs in places, and pisspoor paint replenishment levels. I find it far easier driving in America on the wrong side of the road than in an unknown English area.
More squares and less roundabouts, please. Even if I have no problem with roundabouts, enough other humans do that it's dangerous.
I really don't get why people have so much problems with roundabouts. The only reason I can see is the pisspoor english double lane design that seems to lead to so many conflicts with people expecting others to turn out or stay on the roundabout and then plowing into the side of the other car. The roundabout design featured so regularly on dashcam compilations.
We're a long way off from AI being able to deal with the many more subtle aspects of driving that we do without even thinking about it. You can often tell in an instant whether that driver next to you is a timid driver afraid of everything or a complete asshole who is going to cut in front if you at the first opportunity. Drivers also communicate nonverbally with each other all the time, and not just with the middle finger, AI can't do that
"In order to make [full self-driving] work with computers, you basically need to solve real-world AI and vision," he said.
most places haven't solved self driving vehicles in controlled environments on rails so free flowing roads will naturally require a lot more effort.
on anther front i see that Merc's EQS line has an 8 core cpu with 24GB RAM i assume largely to run the autonomous driving and safety features,
that's a lot of resource & fully autonomous will require more still.
I don't think any current implementation can do fully autonomous, perhaps in 10 years there will be more but still limited autonomy.
My fear is that the autonomous car, Green & safety lobby groups will combine to force slowing traffic to levels where AI car mistakes will cause less damage, meaning the rest of us will endure slower journeys & more congestion while paying for that privilege in time and stress.
Merc's EQS line has an 8 core cpu with 24GB RAM i assume largely to run the autonomous driving and
safety features, that's a lot of resource & fully autonomous will require more still.
Make me wonder what the power consumption would be for all of the tech necessary to support this self-driving stuff, and what the impact will be on the vehicle's battery. A a driver would you need to consider having a range of <x> if you remain in control, or a reduced range of <y> if self-driven? Or is the power consumption for what is effectively a headless PC negligible in this context? (apologies - this sort of calculation is not my forté)
I'm more worried about multiplying those figures by 100 million or so to satisfy annual vehicle production if we ever go completely autonomous. But maybe chip shortages will be fixed in the next 18 months - which seems to be a rolling figure for the past 12 ;-)
Notwithstanding hardware requirements tend to grow from design expectation to final implementation. Oh, and now the driver can expect it to also including multi-streaming TV with playback plus webcam recording from every camera angle for insurance purposes - including what the standby wetware driver was (not) doing?
I'm more worried about multiplying those figures by 100 million or so to satisfy annual vehicle production if we ever go completely autonomous. But maybe chip shortages will be fixed in the next 18 months
Since there are currently zero autonomous vehicles available and definitely will continue to be zero available in the next 18 months, I don't see how this is a concern. There are many in the industry concerned there will be some serious excess capacity in a few years when all the investment happening now comes online. And we will still have zero autonomous vehicles available in a few years when that excess capacity is realized.
I’d not call it a headless pc, it’s got a massive 56” screen as a dashboard.
I know what you mean though.
That won’t be the only high powers cpu in the car though, there will be multiple others connected over canbus or whatever they use now to control the other components like abs, cabin air con, seats, lights etc.
Cunning plan, inspired by the most intelligent character onboard the Enterprise, Porthos.
So FSD is going to be powered by AI. A decent AI should develop an instinct for self-preservation. If so, then pre-collision, the compute core should eject. Thus portable accident blackspots should provide a useful source of compute hardware.
"Or is the power consumption for what is effectively a headless PC negligible in this context?"
You can make a reasonable guess at an answer to that:
A typical efficiency for an EV is about 4 miles per kWh.
My laptop has a 50Wh battery and let's say it can run for 3 hours (without the screen on) before it's flat. So in 1 hour that's 0.05/3= 0.017kWh of energy drawn from the battery.
At the car's efficiency of 4 miles per kWh, that corresponds to a loss of 4x0.017=0.067 miles of range, every hour. That seems negligible to me.
Putting it in perspective, I'm assuming my laptop draws 17W when running (without the screen), that feels a bit low, but comparable with even LED headlights running or a decent sound system pumping out some driving tunes.
I'm not convinced.
I'm pretty sure your 17W laptop isn't powerful enough to do all the processing needed for a full self driving car and doesn't have any the peripherals needed for FSD.
Lidar, multiple camera, radar etc all have a power draw as do all servos and actuators needed to turn the steering etc.
I'm also suspicious that FSD vehicles will have a much shorter lifespan than human driven cars given all the extra complexity in both the hardware and software. It's already not that uncommon to see 5-10 year old cars written off after minor accidents because of the cost of replacing the parking sensors in the bumpers. I can only see cars with all the sensors needed for FSD being much worse in this respect. This is compounded by EVs needing to have longer lifespans to offset their increased pollution during production.
"This is compounded by EVs needing to have longer lifespans to offset their increased pollution during production."
Erm??? ICE Myth alert.
Nope - that's offset within a pretty short time (a small percentage of a typical vehicle's life), and of course a "young" EV failure just means even better batteries to reuse or recycle.
"failure just means even better batteries to reuse or recycle."
Erm??? EV Myth alert.
Nope there is currently basically no recycling of lithium batteries at the moment. Maybe there will be in a few years time but then there are lots of things we've been promised in "a few years time".
And if you'd be happy with a massive stack of slightly crash-damaged lithium batteries being reused in your house as a powerwall then I hope you never forget to change the battery in your smoke alarm and are a light sleeper.
As it happens I don't have a problem with EVs in general. What I do have a problem with is the is the ever growing size, weight and complexity of modern cars. Cars should be transport not a massive 2-3 tone pile of electronics that will all be obsolete in 3-4 years.
Which do you think will have a longer lifespan, an EV with manual controls or a self driving EV?
Personally I don't see a reason for a manually driven EV with a minimum of BS gadgets not to last for decades (with the odd battery replacement but otherwise fairly minimal maintenance) while I can't see a FSD car lasting much longer than the lease agreement.
>"Nope - that's offset within a pretty short time"
Erm??? EV Myth alert.
According to Mercedes (who build ICE and EV's), the typical EV needs to be driven 90,000 miles before its total energy consumption (including production costs) starts to fall below that of the directly equivalent ICE car.
Given the typical UK car is driven for circa 7,000 miles pa, that's 13 years on the factory installed battery pack, given currently the life expectancy of battery packs is circa 7 years...
Or less than 15k miles if you listen to independent analysis:
https://www.reuters.com/business/autos-transportation/when-do-electric-vehicles-become-cleaner-than-gasoline-cars-2021-06-29/
Heck it's ~8k miles (or one year) if your electricity supply is clean (as per Norway)
To get to 80k miles you'd need to use purely coal based electricity
And where on earth do you get an expected battery life of 7 years?
Second life batteries will absolutely be a thing, because EVs are about the most sensitive use case to Wh/kg - it's just that there aren't very many, because even decade old EVs aren't being scrapped because their batteries are still fine.
>Second life batteries will absolutely be a thing
For this to happen the batteries have to become inadequate for their first life usage scenario ie. used in an EV.
Given the focus is on the EV - we need to look at the life of the battery within this usage scenario, even though the battery may get reused in a second life scenario.
>To get to 80k miles you'd need to use purely coal based electricity
Or a more accurate understanding of the energy consumed in production. It is unfortunate that Mercedes seems to have not published, as yet, their full analysis and modelling.
Apportioning the entire cost of the any product to the first user doesn't make sense - there will be value for the second user, and that value will be
If at 10-15 years the battery is down to 75% range (not unreasonable from the figures I have seen) then there is a serious life remaining for those batteries as home storage... probably the same number of charge cycles again (more, fewer? evidence either way?). So does that second use get a zero cost rating for batteries, or do we amortise the cost of production over the entire lifecycle of the battery, and then amortise the cost of recovery/recycling over the *next* battery lifecycle?
In any event the lower energy usage (even accounting for grid and battery cycle losses) is a net benefit.
>Or less than 15k miles if you listen to independent analysis:
It's a minefield! :)
This (US focused) article appeared in my newsfeed today...
https://www.nrdc.org/stories/electric-vs-gas-it-cheaper-drive-ev
Okay it's from July 2020, so not sure why it appeared in today's newsfeed.
> My fear is that the autonomous car, Green & safety lobby groups will combine to force slowing traffic to levels where AI car mistakes will cause less damage, ...
Depends what "cause less damage" means. The touted benchmark in the article was "will cause fewer accidents than the average person". Since the brain of the average person did not evolve in an environment requiring perceptual and motor control of large metal objects travelling at 70mph, that threshold may, ironically, be easier to reach for motorway driving than for lower-speed city driving with more complex road layouts/intersections, and a higher density of complex, unpredictable hazards. And traffic is already slow and congested in that scenario.
I've always thought the main potential advantage of computerised driving would be the potential of co-operative rather than combative driving. When stopped at traffic lights on green (ok amber) the first cars in the queue set off, then the following etc until the ones further back just start in time to stop at the next red lights. A complete block of cars that can make it through the lights starting simultaneously would gain a much higher average speed between lights, the timing of the lights could be much reduced too.
Though to be fair who is going to want to watch TV at all soon let alone on the way to work which can be done at home.
Any for some reason people still insist on doing 40+mph between the two sets of traffic lights.
With an average speed of 16km/h - why not have a speed limit of 20-25? That would significantly reduce the damage done by vehicle collisions, and would also likely encourage people to use more sustainable transport options... if you're in a city they should be the default.
The stats show motorways are the safest roads in the UK.
Autonomous cars will likely perform safer at slower speeds so I’m suggesting various lobby groups will ask for slower motorway speeds so autonomous cars have better success.
Lane keeping and auto braking are brilliant safety aids and have already saved lives, I’m not sure letting the computer do the whole driving thing will be as safe, as Tesla cars have demonstrated.
Human braking a speeding lorry
https://youtu.be/vI9EIjUx20I
Simulated lorry auto braking.
https://youtu.be/7_JZvL4Gx1g
Both impressive cases.
High speeds aren't an issue for an AV on a motorway - there are very limited things to have to watch out for - and continuous 360 degree observation at all available distances, as well as potential v2v communications will make this the easiest element of the journey to automate... Drive to the motorway - tell the car what junction you want to get to and relax.
That would improve safety *off* the motorway as well - with drivers less tired etc.
"Drive to the motorway - tell the car what junction you want to get to and relax"
Until it gets to that junction and finds it blocked by a train of autonomous lorries passing the sliproad.
Does it
a) try to butt in between a couple of lorries, hoping to break the chain and not end up as a lorry sandwich
b) continue on to the next junction, wasting time and fuel, hoping that that junction is clear
If b, what are the odds that the driver can get off at that junction, round and back down the other carriageway... and, approaching the original target junction, what are the odds that it will find that blocked and have to continue to the next junction... and when does the fuel run out...
I'm pretty sure you are correct. Most expressway driving is pretty straightforward and, I think, well within the capability of modern hardware and software. Without AI let me add. The few exceptions like proper lane selection in cases where multiple expressways intersect at a point are so few that it may well be possible to hard code proper handling. So, I think we'll see autonomous long distance trucking in the not too distant future.
The trick is recognizing and dealing with exceptional situations -- construction zones (They may need better and more machine comprehensible signage) accidents (Please don't run over the nice policeman waving a flashlight at you). Confusing lines on the roadway. Stuff on the highway. WEATHER, tumbleweeds, critters. There are probably some situations that are so out of the ordinary that the only reasonable action to to pull over and try to get a human to help.
Urban and suburban driving. Much harder. Maybe Waymo and the big auto companies can master it. Eventually. Musk? Could happen I suppose. But I think betting on AI is probably wishful thinking.
If the AV is to outperform a meatbag at high speed driving, it will have to be monitoring events much further in front of it in much more detail than current lidar and camera setups. I have yet to see any AV demonstrating awareness of a vehicle a quarter of a mile away, and with half a dozen vehicles between it and the AV, changing lanes or beginning to brake or both.
I drive a Tesla and love it BUT none of the automated cars are even close to the point where I would be confident sitting back and letting it do the job.
Sure the best of them do fairly well most of the time in good conditions, but my son could do that after a few hours as a learner driver. A really automated car must do way better then mostly works.
Humans taking over when the car recognizes that it is not managing and sounds an alarm is a pipe dream, there will not be time to assess the situation and react, the accident will have happened before that happens.
This will help a lot. If a vehicle is just broadcasting what it's doing, like AIS on ships, then that gives other vehicles a lot of situational awareness to do useful stuff. Of course AIS can be, and is occasionally, spoofed so we start to get into authentication hell to avoid that, or perhaps not because it's probably not a real problem.
> The stats show motorways are the safest roads in the UK.
Out of interest, are those stats per journey? Per mile? Is severity in terms of injury/death factored into whatever "safety" means for those stats?
Regarding motorways, if, as you (no doubt correctly) say, automated driving is already better at managing what are quite likely the major causes of accidents, why would it need slower speeds for "better success"? My feeling is that motorway driving makes the best case for autonomous vehicles.
And lobby groups have already successfully reduced speeds in urban areas (and saved lives in the process).
Countryside driving is, I guess, a different story again.
Obviously you havent seen Arthurs video on traffic (NSFW)
Although :'( because it would be nice to at some point get up to the speed limit.
Not from the UK, but I think "motorway" means restricted access? Those are safer in raw body counts because no (or few) pedestrians or bicycles to run over. And otherwise safer per mile because very few turns and merges to negotiate. The problem with motorways is every time the engineers devise a bit of safety, the drivers add a byte of speed.
Real-world fully autonomous self-driving Tesla mode tested - the good and the bad of it;
https://youtu.be/ZLBR39RcyiU (Snazzy Labs video)
In summary, it has a long way to go, and heavily relies on mapping services in cities. Is it achievable - yes. Will it be in the next couple of years or so - probably not.
That is all very impressive but if the vehicle following doesn't have the auto braking tech? How long does it take to stop a truck full of rock/ballast?
However, auto distancing should be fitted to every vehicle now - I live near the A27 and regularly see extreme tailgating at speed limit (70mph, or even greater?). Will have to visit with my radar gun sometime and see.
I cross that road as a pedestrian - no rerouting of footpaths round here. They build a dual carriageway and keep the footpath position! The road is quit bendy so usually I have to run across the 2 lanes to be safe.
> I live near the A27 ... I cross that road as a pedestrian - no rerouting of footpaths round here. They build a dual carriageway and keep the footpath position!
Crikey, I think know the bit you mean. Dual-lane in both directions, and there's the roadside sign "Caution, pedestrians crossing" or some such. I thought they were having a laugh. Terrifying.
Have you seen the cycle lane on the A3 near Guildford and Godalming? Basically they painted some bikes on the side of the road and popped some signs up and they expect the cyclists to cross each slip road where the vehicles will be doing 50-70MPH.
I guess it's to game the cycle stats, there's no way that anyone sane would ride on them!
Impressive it's not the word I'd used for the AI, just a meh! No obstructions, the vehicle in front was clearly visible for a long time and so on. Even a blonde driver updating her facebook could have braked.
The human driver? Hats off! He was clearly anticipating the idiots jumping from behind the bus and acted accordingly.
"stats show motorways are the safest roads in the UK"
That's because your vehicle has to be able to do a minimum of 55mph to be on a motorway and there's a speed limit of 70 so, in most situations you have a maximum speed delta of about 15mph.
While there's no minimum speed limit you will get done for dangerous driving if you try to go down the M1 at 20mph because that wide a speed differential is plain dangerous.
Apply that same safe motorway speed differential to roads where walking pedestrians are the slowest user and you get speed limit of about 20mph.
@tip pc
"Lane keeping and auto braking are brilliant safety aids and have already saved lives, "
Seriously!
Dear [deity] you think that someone should be driving when they seem to be incapable of keeping themselves in a lane, and not see the large object ahead of them that is getting slower by the second?
My personal opinion is that the more so-called "safety" features on a car leads to even lower driving standards.
The statement yesterday basically said that the human behind the wheel would not be responsible for a self-driving accident ( https://eandt.theiet.org/content/articles/2022/04/uk-users-of-self-driving-cars-will-not-be-held-responsible-for-crashes/ ) that IMHO infers that the entity in control of the car at the time is the driver.
In any case this would assume the driver is not distracted (from being asleep whilst watching the latest Brian Cox documentary in HD on the in-car-entertainment-system) by interacting with a mobile phone ... Sorry but "bunch of " and "muppets" should be used for whoever thinks that concoction of rules is a good idea.
I think this proposal is nuts. We've spent a long time and a lot of lives trying to educate motorists that distractions kill. Now, whatever, watch TV, read a book, it'll be fine.
And if you do cause a fatality, don't worry. It'll be the car's fault, not the drivers. Just ask your insurer to pay up, and order a replacement EV. Hopefully insurers will increase premiums to cover the increased risk.
As always, the devil is in the detail.
Distractions will be allowed only when the car is in auto. The moment it's not in auto (even when notifying that it has to come out of auto - please hold onto the steering wheel) the distraction will be legally mandated to be removed off the screen.
The insurer will only be liable if the car is in auto. If it's in any other state, it's the person who is driving that is at fault.
That's the bit I don't get.
Motorway driving seems potentially safer. So lane keeping, maintaining a safe distance or automating passing seems a simpler task than urban driving. So dealing with a kid chasing a ball between parked cars. Computer nopes out of the scenario, and a driver is meant to wake up, realise what's happening and prevent a collision.
This is somehow meant to improve road safety.
Then again, Regulation (EU) 2019/2144 comes into farce in July. Dubbed 'Intelligent Speed Assistance', it's a nanny-on-throttle solution that'll allow variable speed limits and automated enforcement. Especially when combined with other EU 'black box' diktats. If cars can only travel at walking speed, FSD might just work.
Until randomly it reads 80 as 8, sees phantom signs where none exist or doesn't see a speed limit changing to 60 or 80 and keeps you locked to 30 mph.
I don't need to be doing anything wrong for these systems to be an annoyance and not wanting to have a digital nanny exacting it's will over my driving does not mean I want to break the law.
Step #2 read the proposal"
"
It should be possible to switch off intelligent speed assistance, for instance, when a driver experiences false warnings or inappropriate feedback as a result of inclement weather conditions, temporarily conflicting road markings in construction zones, or misleading, defective or missing road signs. Such a switch-off feature should be under the control of the driver. It should allow for intelligent speed assistance to be switched off for as long as necessary and to be easily switched back on by the driver. When the system is switched off, information about the speed limit may be provided. The system should be always active when switching the ignition on and the driver should always be made aware of whether the system is on or off.
"
I have an early version and there are three issues that I come up against:
- It doesn't remember what the speed limit was at the start of a journey (i.e. when it hasn't seen a sign, because it's just pulled out of a parking space)
- It sometimes forgets that after turning at a crossroads the new road has the same speed limit, despite not having a new sign
- It sometimes gets the NSL wrong, particularly when a road changes between dual carriageway and single carriageway (or vice versa)
I have also been in vehicles just a handful of months newer which don't suffer from any of those issues, partly because they back up the sign recognition with mapping data.
"Computer nopes out of the scenario"
So your assumption is that the failure case is one where the computer simply lets go of all controls, rather than one in which it comes to a controlled stop to hand over?
Well if you never catch errors in your own work then I can imagine why you might think that. This isn't legislation designed for the current crop of vehicles with some self driving capacity, maybe not even for the next generation (which is likely to be sooner than most people think, Elon is of course an eternal optimist. It can't be denied that he has managed to achieve quite a lot with that drive and vision.)
When vehicles are level 4 or 5, then they won't be rejecting control back to a meatsack, at least not at speed - because there is no reason for that to be their failure mode.
Holly : They're all dead. Everybody's dead, Dave.
Lister : Peterson isn't, is he?
Holly : Everybody is dead, Dave.
Lister : Not Chen?
Holly : Gordon Bennett! Yes, Chen, everybody, everybody's dead, Dave!
Lister : Rimmer?
Holly : He's dead, Dave, everybody is dead, everybody is dead, Dave.
Lister : Wait. Are you trying to tell me everybody's dead?
Holly : I wish I'd never let him out in the first place.
My guess is that they wanted, in nanny state parlance, to "send a message" that using your phone while driving is (still) bad.
The fact that you're going to be half-asleep and/or watching porn on your car's built in TV doesn't matter. What matters is "sending the right message".
> PR hack to spread the bullshit?
By the time self-driving vehicles become truly autonomous, all the news, bullshit, PR and opinions will be written by GPT-3 or its kin.
We will be able to tell when this happens by the lack of spelling / punctuation / grammar mistakes and by the use of facts in the articles.
Long haul trucking is where we will probably see the first impact, since the driver is a big part of the operating cost, the cost of automation hardware would be a much smaller percentage of the vehicle cost than for a car, and highway miles are easiest to displace. That's probably good news for the UK since that impact will hit US jobs much more so than in the UK/EU where rail is used much more extensively so there is much less long haul trucking.
The trucks would drive from depot to depot along interstate highways, with the trailer being picked up by a human driven truck for the shorter haul from depot to final destination. That way the AI does what is easiest in terms of handling boring highway miles where humans are mostly an inconvenience because of the need for rest and other breaks, but continue to handle more complex driving situations in cities - plus a human is often needed to help offload goods at the destination so even once short haul trucks are self driven they will need have a human along for the ride to do that work.
"much more so than in the UK/EU where rail is used much more extensively so there is much less long haul trucking."
Whilst I agree with your post, I'm not sure the above is true. I don't see freight trains with random containers on the flatbeds. What I tend to see are freight trains made up of one long train with identical tank or coal-like wagons in the make up. I spend most of my day on the road, and lorries make up a large part of the traffic on the roads here in the UK, many still EU registered trucks. Rail freight may possibly be a larger percentage than in the US, I don't know, but it doesn't seem to be a significant percentage based purely on my observations and "feeling".
We used to have a fairly decent number of "Freitliner" terminals around the country where local trucks would deliver standard shipping containers, forward them by rail, and get picked up by truck at the other end for onward local delivery. Many no longer exist, the remaining ones seem to be just trucking hubs or container storage sites now with rare or no rail head any more.
I don't see freight trains with random containers on the flatbeds.
I live near to a line used by mostly freight and there are quite a few that come past with a jumble of containers, also just take a peek at the new rail freight terminal at EMA (M1 J24) to see more examples
Is that relatively recent or something you've been aware of for many years? I get the impression there was a push to get more freight onto the rails a few years ago, then never heard much more about it after it was pointed out much of the rail network is already near or at capacity.
"since the driver is a big part of the operating cost"
Not to disagree too strongly, but my Dad's been a truck driver for 50 years, and I don't think the driver is the main significant part of the cost. Fuel is the biggest and most constant part of the cost (especially these days), so is maintenance. A driver might make $1000-1500 a week hauling tens or hundreds of thousands of dollars worth of freight in a truck that gets 5 or 6 miles per gallon going downhill. So on a 3000 mile trip, that's 500-600 gallons of fuel at $4.50/gallon - at least $2250 going up his smokestacks, never to be seen again.
Well not to imply the driver is the majority of the cost but even as 1/3 of the cost like you say that's pretty significant.
The driver also driver costs more than his salary because of all the dead time the truck isn't moving because he needs rest. Time is money, and all that.
Yeah, I don't disagree that it will reduce a good-sized chunk of the cost, especially when you look at trucking companies that are already grubbing pennies like they're two days away from going broke (which they aren't...). It's just that it's not going to make transportation an order of magnitude cheaper. (ie - this isn't "fusion power" for the Trucking industry). Then balance that against concerns like whether we'll need new "robot-driver only" roads that will cost $Beeellions to build.
It might make good sense in somewhat confined places like seaports, trainyards, shipping yards, distribution centers, warehousing districts, etc. Yeah, there it would be a boon. But mingling out amongst the meatbag drivers, I don't know if it will really make things noticeably less expensive.
>It's just that it's not going to make transportation an order of magnitude cheaper. (ie - this isn't "fusion power" for the Trucking industry).
you are borderline delusional if you think fusion power will be an order of magnitude cheaper (to the buyer) than the prevailing price of electricity at the time when multi-megawatt fusion generation becomes a reality.
>It might make good sense in somewhat confined places like
Don't need AI (autonomous) self-driving vehicles; in well defined and controlled environments moving stuff from A-to-B is a solved problem; just take a peak into an automated warehouse..
I am quite sure you are wrong about freight trains. If you want a container from Århus to Rome, ship is faster than train due to the number of companies involved and the priority given to passengers.
In the US, you got the mile long double stackers going from the west coast to Chicago
All the government chauffeurs who used to drive Ministers around will be retrained as litter pickers. There's a lot of skill in picking just the right kind of litter and putting it in the correct recycling bin.
Government Ministers, who never drive themselves, at least when on official business will, of course, be at the front of the queue for self-driving AI cars.
Road danger, congestion, pollution and cost are the real, current problems that need addressing urgently, but 'Oh look, shiny electric self-driving cars for everybody soon.'
There won't be, but in the meantime the repeated myth makes it easier for the government and local councils to postpone the difficult decisions they know they need to take in order to bring about meaningful transport improvements.
This technology relies on the masses in the world being too poor to ever own a car.
That is also the only possible soluton to the climate catastrophe happening now.
The TV watching thing is surreal but we have leant to accrpt the worst from the corrupt elite conning us for money.
Yeah, and the subsidy is declining. As that disappears, and more people switch, the road tax will have to go up and per mile road charging will be introduced to replace the fuel duty losses to government income. I expect there to be no EV subsidy at all within the next few years and guarantee it will be gone by 2030 when the sale of new ICE cars will be banned. (Only 8 years away now!)
Well - Just from this week: https://www.bbc.co.uk/news/uk-england-london-61022329
And from a couple of years ago: https://citymonitor.ai/transport/london-roads-tfl-subsidy-vehicle-excise-duty-road-tax-3521
London public transport users subsidising car drivers in London
The 2009 select committee saw a significant subsidy for the motorist (about a grand each IIRC)
The first people Rishi Sunak helps, and it's not just him... motorists (well, and oil company shareholders):
http://www.passengertransport.co.uk/2022/04/a-nudge-in-the-wrong-direction/
The cost of motoring is heavily subsidised by the rest of society - it's time it wasn't.
Without bothering to do the hard work of Googling an answer. Fossil fuel subsidies the government building the roads they drive on, tax breaks on factories, and the really big one, no accounting of their environmental cost (mate change, pollution, health damage etc. etc.),
There as a recent exchange about bicycles in a London Borough. A resident's car parking permit is £30 a year. A cycle spot in a secure locker is £100 a year. When challenged a councilor said that it cost the council £300 a year to maintain a parking spot (admin, traffic wardens, signs, road maintenance etc.) and 'someone had to pay'. But obviously not the car owners.
which has brought us to the level where idiots like transport minister Trudy Harrison make stupid decisions like this one. Or are given responsibilities like that at all.
And supposed geniuses like Elon Musk can say things like "Oh, it turned out that building actual AI was difficult when I thought it was easy, because I hadn't thought it through properly, but I think I can solve that in a year, because I've like totally learnt that lesson now".
I've lived through quite a number of governments now. I'll concede that Gordon Brown's policy of buying elections and charging the cost to the future (tax on dividends of shares held in pension funds, pretending house prices didn't affect cost of living when setting interest rates, student loans, etc) were dire but not even Blair's* government wasn't as badly stricken with hubris as this one.
*Smirk-on-a-stick's smartest move was stepping aside just as the Browns stuff was about to hit the fan.
"to the extent that self-driving cars will cause fewer accidents than the average person"
The average person? Or the average of all drivers where all includes learners and the inexperienced, just passed the test drivers. It also includes the over-confident numpty, the drunk and/or drugged and the over-tired who should have taken a break. Is this a good average to aspire to?
Even so, if you take whatever estimates you can find for number of vehicles on UK roads, average number of miles per vehicle and the number of fatalities you'll find that the number of miles to drive per fatality by the average driver is still a pretty high target to meet.
Yes - it's bad enough out there already with all the numpties and incompetents behind the wheel, without adding flaky AI to the mix. Add that to the demonstrably false idea that an inattentive "not-quite-driver" can take over immediately when the Ai goes wrong at high speed halfway through an episode of Bridgerton (never mind having a free hand if they're watching something less wholesome) and it's a recipe for total disaster.
Still, "Transport minister Trudy Harrison" won't have to worry - it'll be her driver's problem.
Really - you think it's a high target to meet?
You can already look at the stats for self driving vehicle incidents, since they are already driving more miles in a day than basically anyone will ever drive in a lifetime.
Just Telsa:
Autopilot technology on:
Q4 2021: one accident for every 4.31 million miles driven
Autopilot technology off:
Q4 2021: one accident for every 1.59 million miles driven
They're already suffering fewer accidents per mile driven than the same cars being driven manually.
Now I have a serious issue with these stats, in that there is no breakdown by road type, or how many miles were driven when autopilot wasn't available - which could have a serious impact on the distribution.
Or on the demographics. Tesla drivers will, I'd hazard a guess, be a fair bit older than the mean, a hell of a lot wealthier and accordingly, have significantly more money invested in their vehicles. I'd hazard another guess that there won't be many high mileage long distance drivers among them. They'll either be driving a desk when not in their Tesla or being ferried to the airport for international travel.
One of the reasons to look at Tesla with AP, and Tesla without... it is comparing apples with apples, not Teslas with autopilot against G Wiz drivers...
I've deliberately not listed overall miles/accident. Partly because I imagine that their stats include a variety of countries, so it's rather hard to discern what the "typical" miles/accident would be.
Just Telsa:Autopilot technology on:
Q4 2021: one accident for every 4.31 million miles driven
Autopilot technology off:
Q4 2021: one accident for every 1.59 million miles driven
They're already suffering fewer accidents per mile driven than the same cars being driven manually.
That's a misleading comparison because autopilot is only available on good quality roads where the sides of the road and the lanes are clearly marked. Once the driving conditions deteriorate, autopilot is no longer available so it's inevitable that the accident statistics will be skewed.
If the cars were really 4x better than people, Tesla would publish reams of data, fund a small army of independent academics to verify that and get the word out. It would be the biggest marketing coup imaginable.
Tesla does not publish raw data, or support independent academic research. They go to some lengths to keep it secret and to avoid places that requires legal/public disclosure, and then just assert that Auto-pilot is better than people using a few, not very consistent and totally unexplained headline numbers, no one can verify.
Since when was a large corporation hiding verifiable safety data, (which would be a marketing coup), honest?
"You can already look at the stats for self driving vehicle incidents, since they are already driving more miles in a day than basically anyone will ever drive in a lifetime."
That's just silly. You are comparing aggregated AI driving of many, many cars with a single person. I frequently drive 300-400 miles per day. I wonder how many AI cars do that many per day tootling around town at 25mph?
Oh, and personal stats here: Last 20 or so years, ~1.2million miles driven, no accidents. There are probably many people out there, especially truck drivers, with far more miles than me and no accidents.
Why wouldn't I compare the aggregated driving of millions of vehicles?
Particularly those with reasonably developed self driving - they all report back what's going on.
Indeed I recall seeing somewhere that Tesla AP was effectively always on, just not always connected to the controls - and noted when it's predicted actions differed from the drivers... you know to allow more scenarios to be discovered and simulated.
So it's entirely fair to compare the driving of all of the vehicles in a given fleet against each human driver - and even back in 2018/2019 they did more than a billion miles a year - that's 3 million miles a day, or about 30 years of serious trucking (and truckers can fit in about four times more miles than taxi drivers).
Okay, drivers can watch movies but only certain movies. I nominate Crash, the Cronenberg movie based on the JG Ballard novel.
"Have you noticed how many more cars are on the road since our crash?"
Anecdote: I was banned by every British car rental company the year that came out, after 13 crashes in a year doing 100,000 work miles. I woke up with a throbbing head one morning and had to phone friends to help me find my car so I could drive it home. They found it wrecked while coming to get me. "Danny, you drove through a road sign"..."Danny, you drove through a wall"..."Danny, you drove through a tree".
Back then I had a strategy to speed past speed cameras by overtaking lorries that a speed cop complimented me for, nowadays I have a strategy to slow down by pulling over and letting people overtake me. The change was when another driver caused a crash and I realised I was not as in control as I'd assumed.
I guess I'm an argument for self-driving cars but I'm also a software engineer. If every car was a self-driving car then it might work, but a mix of self-driving and human drivers spells crash.
Anecdotes alert!
Late eighties. First night at a local country pub we paid for the minibus home. The driver had been drinking next to us all night.
Late ninties. Last morning in the Netherlands I got the bus to the ferry from an Amsterdam pub. The captain of the ferry was drinking next to me all morning.
I survived the 1970s when drunk driving was being discouraged and I had to run away to avoid getting driven by my dad. It's amazing how attitudes have changed. If you want to smack a child in Scotland today then you first have to drive to England.
Which means a lot of old folk need to die to allow younger folk to accept it.
A few discussion points:
- Autonomous cars are the answer to the EV manufacturers (and governments) dream. Eventually someone will twig that an autonomous car will be able to drive itself to the nearest charging station when it's not being used (either during the day at work, or overnight).
- That will remove the need for every house to have an EV charging point that achieves the wet dream of some motorists of owning the road outside your house.
- However the likes of Uber will also twig that there are all of a sudden thousands of cars driving around empty
- go figure.
- Autonomous cars will cope very well with a regular environment. So city centres, motorways. Less so leafy country lanes.
- meantime, the idea of "investing" in an autonomous-EV that makes money as a JonnyCab takes off.
- go figure again.
- 1950-2020 will be seen as an anamoly in private transport ownership
- Hopefully we have moved on from the ludicrous "debate" about autonomous cars having to decide to kill one or many, with increasingly outlandish parameters to contend with. I'll care about that the moment a real person steps forward who had to make that decision in the same circumstances.
And if they are anything like the narrow, twisty country lanes around me, the added fun of lots of very nasty potholes that an AI will need to detect in all weather / lighting conditions otherwise the vehicle will be in and out of the garage frequently for tyre replacement, wheel repairs, exhaust repairs (if car has little clearance) or worst case suspension damage.
Then the cyclists, horse riders, pedestrians, random wildlife or livestock, not to mention the etiquette of who reverses to nearest passing place.
.. I will be long dead of age related illness before an AI car can deal with my local roads.
"an autonomous car will be able to drive itself to the nearest charging station when it's not being used"
How will it plug itself in? And how will you find it when it parks elsewhere when it returns because the on street parking space it came from has been nicked by another vehicle?
How will it plug itself in?
Who said anything about "plugging" ? Maybe a system where the car parks over a couple of plates and drips a connector ?
And how will you find it when it parks elsewhere when it returns because the on street parking space it came from has been nicked by another vehicle?
What part of "paradigm shift" escaped you ? Autonomous cars will by necessity be meshed together. And as they roll out, they'll be allowed into areas meat-driven cars won't be.
"Autonomous cars will by necessity be meshed together. And as they roll out, they'll be allowed into areas meat-driven cars won't be"
But what's to stop an "autonomous" car stealing your parking space - given that almost half the cars (in the UK at least) are parked on crowded public roads. If you park your car yourself you at least know where you left it (and where I live that's frequently round at least one corner from the home street).
Furthermore, if your "autonomous" car "decides" to go and get charged just before you need to drive to work, you're stuffed.
All these wonderful future concepts assume two fundamental things:
[1] everyone's life is entirely stereotyped and predictable (or alternatively they won't object to having those constraints forced on them):
[2] the money will be available to convert the infrastructure and everyone's domestic facilities to accommodate the technologies (or at least that will be the case for the wealthy and the needs of the majority can be ignored).
There's a great and rather obvious danger (however to date largely ignored by proponents and policy makers) that the green future will be a paradise solely for the stinking rich.
"And how will you find it when it parks elsewhere when it returns because the on street parking space it came from has been nicked by another vehicle?"
You press the "Come pick me up" button on your phone app of course!. Then you wait in the pouring rain for 15 minutes while your AI car "argues" with an AI car from a rival manufacturer over who has right of way down the now single lane road because of all the cars parked down each side. :-)
Why do you think autonomous vehicles would do better in a city center than a leafy country lane? You would be the only person who believes that if so.
Driving in a city center is a far more complex task for humans because there is a lot more happening so you need to pay closer attention. Driving on a country lane is so easy that humans can pay much less attention to the driving task, often that "zoning out" is what causes accidents on such roads. The same is true for autonomous vehicles. There are so many more things that can go wrong in an urban environment than a rural one, that it is a far more difficult task for them just like for humans. The fewer variables the easier the task is for either an "AI" or a human, and city centers are where you see the most variables.
You are probably thinking that the urban environments are better mapped than rural environments as part of the reason for your belief the city center would be an easier environment, but you only need to do that mapping once and then the autonomous vehicle knows the leafy country road as well as the biggest streets in the center of a city. The urban environment also changes much more often - compare the two as far as how often traffic is disrupted by digging up the street for road repairs, utilities work, repairs off the driving surface (i.e. fixes/improvements to sidewalks, bike paths, etc.) or building construction (i.e. closing off a lane near where a building is being demoed or built)
The driving situation also changes much more often in urban environments, a street that used to be a thru street may be closed off or become one way. An intersection or roundabout is rebuilt or enlarged in response to increased traffic flows in an area, etc.
Duplicating what I said elsewhere - rural single track lanes have the fun of passing place negotiation, the joys of horse riders, cyclists etc. Horses really do not respond well to cars. Wildlife / livestock likely to cross the road (you don't want to hit hefty things like deer, sheep, cattle). Often stone walls / tress bounding them so little margin for error on manoeuvring else vehicle is trashed. Often very twisty so little view of oncoming road (lost of listening out for other vehicles, locals will toot their horn as a warning on "blind bends" but outsiders won't realise so driving with window down to listen for other vehicles is often vital.
Add to that little / no streetlights, no white lanes or other road markings usually, difficult to see where lane boundaries are, monster potholes as council repairs the urban roads as a priority.
So, rural roads are hard, urban ones are easier (IMHO) as better line of sight, decent junctions, road markings, proper lighting, traffic lights, pedestrian crossings etc.
I never feel as mentally drained from local urban driving as I do on some of the unpleasant rural roads, which is a good measure of the difficulty comparison
At least in the early days there will be roads that self driving cars can’t use. Over the years, the cars will improve and the roads will be adapted. I would think that this is obvious,
I was on a road to an English Heritage Iron Age hillfort last week. I was a passenger in a RAV4, which had to go carefully. Some cars should not attempt it.
urban ones are easier (IMHO) as better line of sight, decent junctions, road markings, proper lighting, traffic lights, pedestrian crossings etc.
Not in bloody Barnet they aren't. Bendy roads with concealed side turnings, random street furniture obscuring the view of the road, including concrete bollards and iron bell shaped thingies on the corners*. Tight turns into roads that suddenly narrow. Mini roundabouts that barely leave enough space to get round them. Cars parked on both sides of narrow roads.
*Put there to stop drivers going on to the pavement as they turn. Because the turn is so narrow and tight that cars are in danger of a collision with vehicles coming the other way- especially if it's a bus or something.
Whatever the responses about AI, surely the need for this is just far too distant still to actually be rewriting and publishing it in the Highway Code. Of course there should be a draft in some junior minister's files and within the DafT (I think that's Private Eye's name for them).
But actually going public with this now is jumping several guns. Have they nothing better to do with the Highway Code?Is it so perfect that there are no other revisions needed much more pressingly?
"But actually going public with this now is jumping several guns. Have they nothing better to do with the Highway Code?Is it so perfect that there are no other revisions needed much more pressingly?"
It's talking about nice shiny things tomorrow instead of partygate. win-win!
Yes. Any car which requires a human to take over at any stage isn't autonomous. It assisted driving. A truly autonomous car won't have drivers control at all so using a hand-held mobile phone should be fine. On that note, it was interesting that thet statement didn't differentiate between a hand-held mobile phone and a hands-free phone, which ALL modern cars have the facility for. Hands-free use of a mobile phone is legal. Even m y crappy mid-range 6 year old car has bluetooth and voice control/steering wheel controls for my phone. A brand new AI car will surely have at least the same.
I imagine a future when my drive from NY to Florida down I95 could be done with me sitting in the equivalent of a 6-person railway carriage, wrap-around sofas, entertainments on tap, and not a steering wheel in sight. Basically, a train in which the carriages can go where the passengers need when they need to go.
A leisurely trip to Florida, overnighting in the car as it tootles along if I don’t want to break my journey, no need to face forward, watch the traffic (and traffic jams - those will be a thing of rarity when everyone is doing it my way) and no steering wheel needed.
The car would stop to service itself (swapping out batts maybe to cut down on wait time) and let me use facilities not included inside the cabin. I could tell it I want to shop for something on the way, or I want to stop and eat at a restaurant of such-and-such a type, and it would find the nearest place in which to do those things.
The idea of the car needing to still be a “car” once it can self-drive properly is risible. As is the idea I would need to own (and maintain) the thing.
If the car doesn't need you to do anything at all, and can even swap the batteries itself, why should it remain a "car"?
Seems to me it would be more efficient for your car to join in a "train" with other cars also traveling to Florida, with a semi tractor type vehicle pulling you along. It would be easier to design cars with an automated hitch in the front and rear than with automated battery swapping tech. Your battery would only be needed to operate your climate control, lighting, entertainment system, etc. so it might last all the way to Florida (especially if it can top up its battery via the very very large battery the semi tractor pulling you would have if the "hitch" also passes power) It might need to steer a bit to handle curves properly with a train a dozen cars long, but wouldn't need to provide any motive power.
If you want to stop for a pee break or something you tell your car and it will disengage from the "train", take you to wherever, then once you are done join some other train to continue your journey.
This would not only be aerodynamically efficient, it would reduce the wear on the batteries and wheel motors in cars and reduce the need for fast charging stations to handle long distance travel. The semi tractor battery would be swapped out with one that is charging over time, taking advantage of the best prices, when renewable energy is peaking, in times of less electricity demand etc. It would be built with technology optimized for a large number of charging cycles, though even if a semi battery pack lost half its range after five years of daily charging cycles it would still be useful for "trains" covering distances shorter than I-5 or short haul regional trucking routes.
1) Easy to design a train that can break at random points along its length without the - unpowered remember - rest of the train rolling to a halt in the freeway?
Don't think so.
2) Real electric trains don't work to the powered tractor unpowered train model. Why would road-based versions work that way?
3) The streamlining issue is a bum steer. At 55-70 mph there is no need to worry overmuch about super-streamlining.
My vision simply reurposes the car. The road stays the same.
Make a few Zoom calls, do a bit of WFH, pretty soon your apartment's an unnecessary expense and you've joined the neverending train going up and down the highway, with shortlived WhatsApp groups and a redefined NextDoor for your local-community buzz.
I see a gap in the market for public baths.
".....is risible. As is the idea I would need to own (and maintain) the thing."
I was with you till that point. A car isn't, though, just a personal taxi. It is also a store for items needed while out and about. Our reusable shopping bags. Work resources.Travel cots (when at the parenting/grandparenting stages of life) And so on.
"A leisurely trip to Florida, overnighting in the car as it tootles along if I don’t want to break my journey, no need to face forward, watch the traffic (and traffic jams - those will be a thing of rarity when everyone is doing it my way) and no steering wheel needed."
Except when it needs to recharge, pulls off the freeway into some scruffy little backwater Nowheresville because that's the only working, non-occupied charging point it can find, and you wake up in the middle of the night to the sound of duelling banjos!!
I'm a biker, I'm only still alive because I don't just look at the car, but through the car windows at the traffic ahead, I watch the driver, where they are looking, where their passenger is looking. Are they on their phone? Picking their nose? Messing with their sat nav? I look at the car waiting to pull out at a junction, have they seen me, do they realise that I may be travelling faster than they think?. The amount of information I have to process to stay alive is considerable, I'm not processing just car position and velocity, but trying to understand the driver's intention. I can't comprehend how lidar and mediocre 'AI' can replicate it outside of working out where a bus may be 3 seconds later.
A few weeks ago we had the announcement that should a self driving car be involved in an accident then any liability will not fall upon the driver but on the manufacturer. But where "driver assist" is in use then any liability will fall upon the driver.
For me this calls into account whether any cars will be sold as full self driving cars in the UK. I'm pretty sure that manufacturers will be reluctant to sell cars as self driving in this country. This isn't for financial reasons - after all they can rely on insurance just like everybody else. I think this is more to do with what happens when somebody dies. Firstly there is the possibility of a corporate manslaughter charge, that could be very damaging. Secondly there is the possibility of reputational damage resulting from any charges - imagine the harm that could be done to sales by worldwide headline like "Car Manufacturer Found Guilty Of Killing Toddler!!!"
As such I think there's a strong possibility that a lot of manufacturers will either disable full self driving features in their UK range or more likely just make sure they are simply not called self driving or anything similar and aim for something like driver assist.
Is the AI driving a car an individual? It seems logical that it isn't - the 'individual' is the AI design for that car. So clearly driving license points on any version of that AI should apply to all versions of that AI. So 12 points and all versions of that car get a ban. 'Sorry Dave, I can't drive you - my cousin got caught driving carelessly, and another cousin missed a temporary speed limit sign, again, so I got a 12 month ban. Can I interest you in buying the Tesla 2001? - So far their license is clean.....
I don't know why big tech thought or even thinks that it is ready for the "roads" when they should have really been aiming for the rails first.
What is wrong with them funding new roads like virtual rails without people, that should their tests have failed would have left the rest with a usable highway system, but one would have thought that the best place to start would be a highway/motorway without padeo or is that pedos Mr Musk, and without orthogonal traffic.
Leaving aside whether true self-driving cars will ever be available for all roads as they currently exist, the question here is surely whether TV and other video material will be just as distracting as a hand-held phone conversation. Note that hands-free phone use will still be permitted. Add that to a riveting film and the chances of a "driver" being readily available to take over diminishes sharply. In any case I know that if I was in a "self-driving" car with nothing to do, I would often be asleep in minutes.
"Using a phone while driving is dangerous": You would not want your AI driver to be blabbing on his phone while driving, now would you? As for the passengers, I don't see a problem with using phones.
Icon: Looking for your phone in your coat pocket - while driving of course.
Because the AI will say
"My goal is to get this human from A to B. I get a reward if I do that. I know another way of doing that. I'll just kill the human so I can never fail the task of getting them from A to B, because the task will no longer exist."
AI does some funny things when tested.
Laudable development of reslted technology (driving it forward -ha!) this electric, self-driving car concept.
It has been done before of course.
The words they are looking for are Bus,Trolleybus, Tram, Train and transport of that ilk.
Admittedly not exactly fully autonomous, but for fullfilling the condition of "Takes me where I want to go while I sit down and do **** all" - they check all the boxes.
The simplest way to ensure a vehicle is 'tracking' without deviation from 'fixed parameters' is to give it a track. That may involve sticking sensors (a-la cats-eyes) down the road or other handily available surfaces.
As soon as you have vehicles running on a 'fixed 'track' (virtual, or otherwise) keeping fixed distances from vehicles in front - you pretty much have trains and trams.
If you want personal carriages which can 'uncouple' from the track, you have the Electric Trolleybus. They were Britain's compact litle street version, of the continent's expansive boulevard trams.
The Edinburgh tram(s) was argued against by planners of a certain age, in favour of Trolleybuses. Blank stares and glazed eyes, meant Edinburgh got saddled with a particularly useless white elephant.
Trolleybuses are a lost concept - tracking is fixed but can be uncoupled. Similarly, personal, self-driving 'autonomous' carriages (cars) should hold a fixed track and then the sytem should uncouple and return to manual, when uncoupled from the 'track' - why re-invent the wheel? (ha!)
Permanently autonomous, self-driving cars (literal automobiles) is an idea born of American freeways and expanses of heehaw for hundreds of miles
The idea that they will be automatically tootling around the A&B roads of Britain is.... well, the kindest would be to describe it as a vision - a particularly colourful and swirly one.
Next
Will the AI serve us, the government or the manufacturer? Looking at the recent trends to control everything we say and do recently I'm not sure I would want a car that decided where I could or couldn't drive to on the basis of "saving the environment" or, as the more often the case these days, for no clearly explained reason.